Data Science MODULE 5

Question 1

In machine learning we describe the learning of the target function from training data as

Answer 1

Inductive learning

Question 2

What is feature selection?

Answer 2

Methods, employed to reduce the amount of input variables to those that are believed to be most useful to a model

Question 3

Unsupervised feature selection?

Answer 3

Ignores the target variable

Question 4

Supervised feature selection

Answer 4

Use the target variable in the selection process

Question 5

Wrapper feature selection

Answer 5

Gaan basies en gebruik verskillende inputs, uit jou data, om die beste fit te bepaal

Question 6

Filter feature selection

Answer 6

Soos ek dit sien, basies korrelasies tussen individuele features en die response. Die bestes word gebruik om die model te train

Question 7

Derde tipe feature selection?

Answer 7

Hulle noem dit intrinsic - so n tree based model is baie goeie voorbeeld van dit

Question 8

Is dimensionality reduction n feature selection metode?

Answer 8

Eintlik nie, want nuwe features word eintlik geskep, vanaf die oorspronklike inputs

Question 9

Drie tipe categorical data

Answer 9

Nominal (r,g,b)
Ordinal (1st, 2nd, 3rd)
Boolean (true and false)

Question 10

Filter feature selection
Numeries - numeries

Answer 10

Pearson
Spearman

Question 11

Filter feature selection
Numeries - kategoriee

Answer 11

Anova
Kendall

Question 12

Filter feature selection
Kategoriee Kategoriee

Answer 12

Chi-squared
Mutual info

Question 13

Verskil tussen Pearson en Spearman?

Answer 13

Pearson vir lineer
Spearman vir nie-lineer

Question 14

Verskil tussen Anova en Kendal

Answer 14

Anova - lineer
Kendall - nie lineer

Question 15

Scikit libraries vir:
Pearson
ANOVA
Chi squared
Mutual info

Answer 15

f_regression()
f_classif()
chi2()
mutual_info_classif() en mutual_info_regression()

Question 16

Libraries in SciPy

Answer 16

kendaltau
spearmanr

Question 17

Selection methods in python

Answer 17

SelectKBest
SelectPercentile

Question 18

Using regularisation to constrain a neural network amounts to modifying the

Answer 18

Objective function by adding a penalty term?

Question 19

So in baie eenvoudige terme, wat is regularisation?

Answer 19

Wanneer die weights te groot raak, word dit gepenaliseer

Question 20

Hoe word die mate van regularisation beheer?

Answer 20

Met n scaling factor - lambda

Question 21

L1 regularisation?

Answer 21

Is geneig om weights na nil toe te vat, en konneksies te breek

Question 22

L2 regularisation staan ook bekend as?

Answer 22

Weight decay
Groot gewigte word meer gepenaliseer as kleiner gewigte
Hoe groter die faktor is, hoe nader sal die gewigte beweeg aan nil

Question 23

Tussen L1 en L2, watter een is computationally meer effektief?

Answer 23

L2

Question 24

L1 is nie ondersteun deur die MLPRegressor nie

Answer 24

Korrek, gebruik sknn.mlp as jy dit nodig kry

Question 25

So met cross validation, en neurale netwerke, watter hyperparameter word verander?

Answer 25

Lamda, of ook dan bekend as die scaling factor